17. Investigating Predicted Floor Response Sensitivity to Input Forcing Function Variables

Footfalls are often the most critical source of vibration on elevated floors. Advanced modeling techniques consider walker weight and cadence, harmonics/subharmonics of footfall loading, dynamic loading factors (DLFs), and a resonant build-up factor. Design guides have traditionally treated the pedestrian weight, walking frequency, and DLFs as deterministic parameters; however, recent research has suggested that a stochastic approach is more appropriate. These stochastic models use Monte Carlo simulations to develop probability distributions of the vibration response from thousands of combinations of walking speeds, weights, and harmonics/subharmonics. Unfortunately, this methodology drastically increases computation time, significantly limiting its applicability. In this study, the sensitivity of the floor response to the distribution of possible input variable combinations is considered. Response predictions are developed using finite element models of two actual buildings, with 50 modes of vibration used for each floor. The goal is to determine the variables which most strongly influence the amplitude of the floor response. Using this information, only the important variables can be modeled in future simulations, which will dramatically decrease computational time. These results can be used to develop stochastic models that are both robust and practical.